Skip to main content
gtbsf81
Participant
Oct 21, 2026 3:41 AM UTC
Answered

C2PA AI Metadata Persisting Despite Content Credentials Being Disabled

Product: Adobe Lightroom Desktop (Cloud-Based), Version 9.6
Operating System: Windows 11
Camera Equipment: Nikon Z8 and Nikon Z6 III
Issue: AI-related C2PA metadata remains embedded in exported JPEG photographs despite disabling Apply Content Credentials.

Description of the Problem

I am reporting an issue with Adobe Lightroom Desktop 9.6 involving the automatic inclusion of C2PA Content Credentials in exported JPEG photographs.

My original images are authentic photographs captured using Nikon digital cameras. During normal post-processing, I occasionally use Lightroom's Generative Remove tool to eliminate minor distractions, such as unwanted shadows, lines, or small objects.

These are genuine camera-captured photographs, not images created entirely using generative AI.

Despite explicitly disabling Apply Content Credentials during export, Lightroom continues to include C2PA metadata identifying AI involvement in the resulting JPEG files.

This behavior creates problems when photographs are published on social media platforms, particularly X, where the images are subsequently labeled "Made with AI."

Steps to Reproduce

  1. Import an original Nikon NEF (RAW) photograph into Lightroom Desktop 9.6.

  2. Make standard photographic adjustments.

  3. Use Generative Remove to eliminate a minor shadow, line, or distraction.

  4. Export the photograph as a JPEG using Custom Settings.

  5. Ensure Apply Content Credentials is unchecked.

  6. Examine the exported JPEG using a C2PA verification tool.

  7. Observe that the exported file still contains AI-related C2PA provenance information.

Actual Result

The exported JPEG retains C2PA metadata declaring AI involvement, even though the user has explicitly disabled Apply Content Credentials.

When uploaded to X, these photographs may receive a prominent "Made with AI" label.

Expected Result

Adobe should provide clear documentation and export controls explaining when AI-related Content Credentials are mandatory or automatically applied.

If disabling Content Credentials does not apply to photographs edited with Generative Remove, Lightroom should clearly communicate this limitation before export.

Adobe should also distinguish between:

  • Fully AI-generated images.

  • Authentic camera-captured photographs with minor AI-assisted retouching.

  • Photographs edited exclusively using traditional, non-generative tools.

Business and Professional Impact

This issue affects professional photographers who use Lightroom for routine photographic corrections.

A genuine photograph that has undergone a minor generative edit may be presented to viewers as though the entire image was artificially generated.

This can damage the photographer's credibility, undermine confidence in authentic photography, and create confusion about the image's origin.

Requested Action

I request that Adobe:

  1. Investigate and document why AI-related C2PA metadata remains present when Apply Content Credentials is disabled.

  2. Clarify whether this behavior is intentional, required by policy, or a software defect.

  3. Provide more precise provenance descriptions that distinguish authentic photography with limited generative retouching from entirely AI-generated imagery.

  4. Improve Lightroom's export interface so photographers understand exactly which Content Credentials will be included.

  5. Coordinate with social media platforms to prevent misleading interpretations of AI-assisted photographic edits.

I would appreciate a technical explanation and a formal response from Adobe's Lightroom development team.

    Claire H.
    Claire H.Correct answer
    Community Manager
    Oct 21, 2026 4:58 PM UTC

    Hi ​@gtbsf81, Thank you for the detailed report. This is actually by design, not a bug. When you use a qualifying generative AI feature in Lightroom, such as Generative Remove, Adobe automatically attaches C2PA metadata indicating AI involvement. This is a separate, mandatory process driven by legal requirements (including California SB-942 and the EU AI Act), and it cannot be turned off via the "Apply Content Credentials" export toggle.

     

    That toggle controls the optional Content Credentials (creator identity, edit history, etc.). But the GenAI disclosure layer is applied independently whenever a generative tool is used in the edit.

     

    Workaround: If you want to avoid the "Made with AI" label on social platforms, use the non-generative Heal tool instead of Generative Remove for minor image retouching.

    For more details, see:

    Your feedback about better distinguishing "fully AI-generated" images from "camera photos with minor AI retouching" is appreciated. Please also submit this as a feature request so the product team can track demand for more granular labeling. ^CH

    View reply

    3 replies

    talented_observerb1e9
    Participant
    Nov 3, 2026 7:30 PM UTC

    Disclosure: I make PhotoCleanr, a paid tool. I mention it only for the last case below, which Adobe’s own workaround doesn’t cover.

    As Adobe says above, Generative Remove adds a separate generative-AI record to the export, and the “Apply Content Credentials” switch doesn’t control it. X and Instagram read that record and can show their AI label. They can’t tell a dust-spot fix from a generated image from the record alone.

    • New edits: use the non-generative Heal tool for the retouch, as Adobe suggests, then export.
    • Check what a file says: a Content Credentials verifier shows whether the AI record is in the exported JPEG and which app wrote it.
    • A JPEG you already exported: you can post a copy without those records. ExifTool can remove them, and I make PhotoCleanr, which does it without re-encoding the picture. Its free report on photocleanr.com/guides/facebook-ai-info-label shows what a file contains first.

    This is for a real photo with a minor retouch. If generative AI changed what the picture shows (a person, a place, an event that looks real), the record should stay. Cleaning a file also doesn’t change a post that’s already up, and a platform can still label an image that carries an invisible watermark.

    Translate
    Claire H.
    Community Manager
    Claire H.Community ManagerCorrect answer
    Community Manager
    Oct 21, 2026 4:58 PM UTC

    Hi ​@gtbsf81, Thank you for the detailed report. This is actually by design, not a bug. When you use a qualifying generative AI feature in Lightroom, such as Generative Remove, Adobe automatically attaches C2PA metadata indicating AI involvement. This is a separate, mandatory process driven by legal requirements (including California SB-942 and the EU AI Act), and it cannot be turned off via the "Apply Content Credentials" export toggle.

     

    That toggle controls the optional Content Credentials (creator identity, edit history, etc.). But the GenAI disclosure layer is applied independently whenever a generative tool is used in the edit.

     

    Workaround: If you want to avoid the "Made with AI" label on social platforms, use the non-generative Heal tool instead of Generative Remove for minor image retouching.

    For more details, see:

    Your feedback about better distinguishing "fully AI-generated" images from "camera photos with minor AI retouching" is appreciated. Please also submit this as a feature request so the product team can track demand for more granular labeling. ^CH

    Translate
    gtbsf81Author
    Participant
    Oct 22, 2026 6:06 PM UTC

    Thank you for clarifying that this behavior is intentional rather than a software bug.

    However, your response raises several important concerns about Adobe's implementation and its impact on authentic photography.

    First, I understand the importance of AI transparency. My concern is that a genuine photograph captured with a Nikon camera, with only a minor shadow or distracting line removed using Generative Remove, can subsequently be labeled by X as "Made with AI."

    That label creates a materially misleading impression about the origin and authenticity of the photograph.

    Second, you cite California SB 942 and the EU AI Act as requiring this disclosure. However, Article 50(2) of the EU AI Act specifically provides an exception for AI systems performing assistive functions for standard editing or edits that do not substantially alter the original input or its meaning.

    Can Adobe explain why minor photographic corrections using Generative Remove are categorically treated as requiring mandatory AI disclosure, regardless of the extent of the edit?

    Third, the C2PA disclosure appears to create an unintended consequence. Social media platforms such as X may interpret Adobe's AI involvement declaration as justification for labeling the entire photograph "Made with AI."

    That is not an accurate representation of a photograph captured at an actual event with a professional camera.

    Finally, I have observed that the presence or absence of C2PA metadata can affect whether X displays its AI label. This raises additional concerns about the consistency and reliability of provenance-based labeling.

    I would appreciate Adobe addressing three specific questions:

    1. Does Adobe distinguish between a fully AI-generated image and an authentic photograph with minor generative retouching in its C2PA declarations?

    2. How does Adobe interpret the standard-editing exception under Article 50(2) of the EU AI Act?

    3. Is Adobe working with X and other social media platforms to ensure Content Credentials are interpreted accurately, without misrepresenting genuine photography?

    I support transparency about generative AI. However, transparency must also be accurate, proportionate, and meaningful.

    A photograph taken with a Nikon camera and lightly retouched with AI is not the same as an image generated entirely by AI.

    I would appreciate a substantive response from Adobe's Content Authenticity or Lightroom engineering team rather than a general explanation of why C2PA metadata is attached.

    Translate
      Terms & ConditionsAccessibility statement
      Using the community Terms of use Privacy Cookie preferences Do not sell or share my personal information ad choicesAdChoices